Benatar et al
1
1
Title: Validation of Serum Neurofilaments as Prognostic & Potential Pharmacodynamic Biomarkers for ALS Authors: Michael Benatar MD PhD 1, Lanyu Zhang MS 1, Lily Wang PhD 1, Volkan Granit MD 1, Jeffrey Statland MD 2, Richard Barohn MD 2, Andrea Swenson MD 3, John Ravits MD 4, Carlayne Jackson MD 5, Ted M Burns MD 6, Jaya Trivedi MD 7, Erik P Pioro MD PhD 8, James Caress MD 9, Jonathan Katz MD 10, Jacob L McCauley PhD 11, Rosa Rademakers PhD 12, Andrea Malaspina MD PhD 13, Lyle W Ostrow MD PhD 14, and Joanne Wuu ScM 1, on behalf of the CReATe Consortium Author Information 1 University of Miami, Miller School of Medicine, Miami, FL, United States 2 Kansas University Medical Center, Kansas City, KS, United States 3 University of Iowa, Iowa City, IA, United States 4 University of California San Diego, San Diego, CA, United States 5 University of Texas Health Science Center San Antonio, San Antonio, TX, United States 6 University of Virginia, Charlottesville, VA, United States 7 UT Southwestern Medical Center, Dallas, TX, United States 8 Cleveland Clinic, Cleveland, OH, United States 9 Wake Forest School of Medicine, Winston-Salem, NC, United States 10 California Pacific Medical Center, San Francisco, CA, United States 11 John P Hussman Institute for Human Genomics, Miami, FL, United States 12 Mayo Clinic Jacksonville, Jacksonville, FL, United States 13 Blizard Institute, Queen Mary University of London, London, United Kingdom 14 Johns Hopkins University, Baltimore, MD, United States Title character count: 95 Abstract word count: 233 Word count: 3,029 Reference count: 27 Tables: 3 Figure: 2 Supplemental Data: Table e1. Figure e1 Corresponding Authors:
Name: Michael Benatar, Joanne Wuu
Address: 1120 NW 14th Street, Suite 1300, University of Miami, Miami, FL, 33136
Email: [email protected]; [email protected]
Statistical Analysis: Lily Wang PhD, Lanyu Zhang MS, Joanne Wuu ScM (all affiliated with the University of Miami) Disclosure of financial relationships:
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
2
2
Michael Benatar reports grants from National Institutes of Health, the ALS Association,
the Muscular Dystrophy Association, the Centers for Disease Control and Prevention,
the Department of Defense, and Target ALS during the conduct of the study; personal
fees from Mitsubishi Tanabe Pharma, AveXis, and Genentech, outside the submitted
work. In addition, Dr. Benatar has a provisional patent entitled ‘Determining Onset of
Amyotrophic Lateral Sclerosis’. Dr. Benatar also serves as a site investigator on clinical
trials funded by Biogen and Orphazyme.
Lanyu Zhang reports no disclosures.
Lily Wang reports grants from the National Institutes of Health during the conduct of the
study.
Volkan Granit reports no disclosures.
Jeffrey Statland reports grants from National Institutes of Health, MDA, and the FSH
Society during the conduct of the study; and personal fees from Fulcrum, Acceleron,
Strongbridge, Sarepta, and AveXis, outside the submitted work.
Richard Barohn reports grants from the National Institutes of Health during the conduct
of the study; grants from FDA/OOPD , the ALS Association, the Muscular Dystrophy
Association, Sanofi/Genzyme, Biomarin Pharmaceuticals, IONIS Pharmaceuticals, Teva
Pharmaceuticals, Cytokinetics Pharmaceuticals, Eli Lilly and Company , PTC
Therapeutics, Orphazyme, Neuraltus Pharmaceuticals, Alexion Pharmaceuticals,
Sarepta Therapeutics, and The Marigold Foundation outside the submitted work;
personal fees from NuFactor, Momenta and Plan 365; and other support from Novartis
Pharmaceuticals, Option Care, Platform Q Health Education, Ra Pharma. In addition,
Dr. Barohn has patents on the MG-ADL and QMG with royalties paid.
Andrea Swenson reports grants from the University of Iowa during the conduct of the
study and other support from Amylyx Pharmaceuticals, FlexPharm and Cytokinetics
outside the submitted work.
John Ravits reports personal fees from MT Pharma and AveXis, outside the submitted
work and serves as a site investigator on a clinical trial funded by Biogen.
Carlayne Jackson reports grants from the National Institutes of Health, Cytokinetics and
Mitsubishi Tanabe Pharma America during the conduct of the study; and personal fees
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
3
3
from Mallinckrodt, Brainstorm, Mitsubishi Tanabe Pharma America, ITF Pharma, and
Anelixis outside the submitted work.
Ted M Burns reports no disclosures.
Jaya Trivedi reports grants from National Institutes of Health, Sanofi-Genzyme and CSL
Behring during the conduct of the study; as well as personal fees from CSL Behring
outside the submitted work.
Erik P Pioro reports grants from National Institutes of Health and Center for Disease
Control during the conduct of the study; grant support from the ALS Association and
CDC/NIH outside the submitted work; as well as personal fees from Avanir
Pharmaceuticals, Inc., Biohaven Pharmaceuticals, Inc., Cytokinetics, Inc., ITF Pharma,
Inc., MT Pharma America, Inc., and Otsuka America, Inc.
James Caress reports grants from National Institutes of Health, Amylyx
Pharmaceuticals Inc, Orion Corporation, Cytokinetics Inc, Flex-Pharma and Neurology
Clinical Research Initiative TCZALS-001 outside the submitted work.
Jonathan Katz reports no disclosures.
Jacob L McCauley reports no disclosures.
Rosa Rademakers reports no disclosures.
Andrea Malaspina reports no disclosures.
Lyle W Ostrow reports grants from The Target ALS Foundation and the ALS
Association during the conduct of the study.
Joanne Wuu reports grants and funding from National Institutes of Health, the ALS
Association, the Centers for Disease Control and Prevention, the Department of
Defense, and Target ALS during the conduct of the study.
Running Title: Validation of serum neurofilaments as ALS biomarkers Search Terms: [177] Anterior nerve cell disease; [178] Amyotrophic lateral sclerosis; [21] Clinical trials Methodology/study design; Biomarkers
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
4
4
Abstract (233 words)
Objective. Identify preferred neurofilament assays, and clinically validate serum NfL and
pNfH as prognostic and potential pharmacodynamic biomarkers relevant to ALS therapy
development.
Methods. Prospective, multi-center, longitudinal observational study of patients with
ALS (n=228), primary lateral sclerosis (PLS, n=20) and progressive muscular atrophy
(PMA, n=11). Biological specimens were collected, processed and stored according to
strict standard operating procedures (SOPs). Neurofilament assays were performed in a
blinded manner by independent contract research organizations (CROs).
Results. For serum NfL and pNfH measured using the Simoa assay, missing data (i.e.
both technical replicates below the lower limit of detection (LLD) was not encountered.
For the Iron Horse and Euroimmun pNfH assays, such missingness was encountered in
~4% and ~10% of serum samples respectively. Mean coefficients of variation (CVs) for
pNfH in serum and CSF were ~4-5% and ~2-3% respectively in all assays. Baseline NfL
concentration, but not pNfH, predicted the future ALSFRS-R slope and survival.
Incorporation of baseline serum NfL into mixed effects models of ALSFRS-R slopes
yields an estimated sample size saving of ~8%. Depending on the method used to
estimate effect size, use of serum NfL as a pharmacodynamic biomarker, instead of the
ALSFRS-R slope, yields significantly larger sample size savings.
Conclusions. Serum NfL (Simoa assay) may be considered a clinically validated
prognostic biomarker for ALS. Both serum NfL and pNfH (using the Simoa assay) have
potential utility as pharmacodynamic biomarkers of treatment effect.
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
5
5
Introduction
Therapy development for amyotrophic lateral sclerosis (ALS) is challenging for many
reasons, with the design and interpretation of phase II studies especially so 1. These
“mid-development” studies typically employ traditional clinical measures such as
survival or the rate of decline of the Revised ALS Functional Rating Scale (ALSFRS-R)
as the principal measures of therapeutic effect, but are typically underpowered, making
it difficult to decide which experimental therapeutics to advance from phase II to phase
III 2. Biomarkers with prognostic and potential pharmacodynamic utility have great
potential to help overcome these challenges 2, 3. Controlling for prognostic biomarkers,
for example, might reduce phenotypic heterogeneity, thereby improving statistical power
for a fixed sample size. Similarly, pharmacodynamic biomarkers may help to verify
target engagement or demonstrate the presence of the intended biological effect.
Neurofilaments – both light (NfL) and phosphorylated heavy (pNfH), in blood and
cerebrospinal fluid (CSF) – have been proposed as potential ALS biomarkers with
diagnostic value4-8, utility in predicting prognosis5-7, 9-14, and as possible
pharmacodynamic biomarkers15, 16. The potential diagnostic utility aside, which is not
the focus of this paper, the precise role of neurofilaments as prognostic or
pharmacodynamic biomarkers in ALS trials is not yet fully defined. Several questions
remain unresolved: Which pNfH assay, among the many available, should be selected?
Do NfL and pNfH convey the same information, or should both be measured?
Moreover, understanding the prognostic value of neurofilaments has been hampered by
the use of estimated rather than measured rates of disease progression through
longitudinal follow-up; and inconsistent consideration of the value added by
neurofilaments to clinical predictors of prognosis (age, gender, bulbar onset, etc.).
Similarly, studies of the longitudinal trajectories of neurofilaments have reached
inconsistent conclusions about the relationship with clinical measures of disease
progression, driven in part by relatively small numbers of participants with fast
progressing disease.
Here we present NfL and pNfH data (obtained from serum and CSF) from a large cohort
of patients with ALS and related disorders who have undergone careful longitudinal
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
6
6
clinical phenotyping along with serial collection of biological samples. Serum and CSF
neurofilaments have been quantified using a variety of assays, performed by CROs
blinded to participant identity and status. Combined with carefully collected phenotypic
data, we have endeavored to address unanswered questions related to the prognostic
and potential pharmacodynamic utility of neurofilament levels.
Materials and Methods
Study Population
Patients with ALS and related disorders were enrolled at multiple centers, through the
Phenotype-Genotype-Biomarker study of the Clinical Research in ALS and Related
Disorders for Therapeutic Development (CReATe) Consortium. Rigorously standardized
clinical assessments and biological sample collections were performed every 3-6
months. Samples included in this experiment were from study visits that took place
between April 2015 and November 2017. The University of Miami IRB (the central-IRB
for CReATe) approved the study; all participants provided written informed consent.
Sample Collection, Processing and Storage
Blood was collected in serum-separating BD-vacutainers and allowed to clot upright at
room temperature for 1-2 hours. Following centrifugation (1750g for 10 minutes at 4°C)
serum was aliquoted into cryogenic sterile freestanding conical microtubes (Nalgene or
Bio Plas Inc.) and stored at -80°C. CSF was collected in polypropylene tubes,
centrifuged (1750g for 10 minutes at 4°C), aliquoted into polypropylene cryogenic sterile
freestanding conical microtubes, frozen within ~30 minutes of collection, and stored at -
80°C.
Neurofilament Quantification
Serum and CSF neurofilament concentrations were quantified by Quanterix using their
Simoa NfL17, 18 and pNfH assays19; each plate contained calibrators (0-500 pg/mL for
NfL and 0-2000pg/ml for pNfH) and quality controls. Samples were diluted 1:4 to fall
within the range of the standard curve. Serum and CSF pNfH were also quantified by
Iron Horse Diagnostics, using their in-house assay (which uses monoclonal capture and
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
7
7
polyclonal detection antibodies)10 and the Euroimmun CE marked ELISA20 (which
reverses the capture and detection antibodies). All samples were measured in duplicate
at the same dilution, and all assays were performed blind to clinical parameters.
Statistical Analysis
Symptom onset was defined as the first occurrence of limb weakness, dysarthria,
dysphagia, or dyspnea. ∆FRS was calculated by subtracting baseline ALSFRS-R from
48 (maximum ALSFRS-R score), divided by time (months) from symptom onset to
baseline 21. Survival duration was defined as time from symptom onset to PAV (≥22
hours/day non-invasive ventilation), tracheostomy or death.
Each sample was assayed twice to generate two replicates on each platform. We then
computed (a) the number of samples for which neurofilament levels were below the
LLD, (b) the mean difference and 95% confidence interval of the differences between
replicates (i.e. limits of agreement22), and (c) the CVs between replicates. The
association between serum and CSF concentrations was assessed using Spearman’s
rank correlation. These analytic characteristics were qualitatively compared between
assay platforms and guided our selection of one of the pNfH assays for further
investigation.
To evaluate the prognostic utility of serum NfL and pNfH, we fitted a linear model to
estimate the change in ALSFRS-R for each subject over time, using only the subset of
participants with at least three ALSFRS-R scores. The associations between ALSFRS-
R slope estimates and pre-specified clinical and baseline neurofilament predictor
variables were investigated using multivariable regression. To assess the joint
contribution of NfL and pNfH, a likelihood ratio test was used to compare the full model
(including clinical and neurofilament variables) to a reduced model (including only
clinical variables). The prognostic value of serum neurofilaments was also evaluated by
Kaplan-Meier (univariate) and Cox regression (multivariate) methods, using PAV- and
tracheostomy-free survival as the outcome.
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
8
8
We then estimated trajectories of changes in neurofilament levels over time using a
linear regression fitted separately to data from each subject. This model, which was
limited to subjects with at least three available neurofilament values, included
neurofilament as the outcome variable and time as predictor variable. Because
neurofilament values were heavily skewed, natural logarithm-transformation was used
to help achieve normality.
To assess the impact of considering baseline serum NfL on the sample size for a
placebo-controlled clinical trial with ALSFRS-R slope as its primary outcome, we
conducted a simulation similar to that described by Kuffner23. Briefly, we compared two
mixed effects models, typically used for assessing treatment effects on ALSFRS-R
slopes in clinical trials: Model-1 included ALSFRS-R scores as the outcome variable,
time, time x group (treatment vs. controls), log(baseline NfL), time x log(baseline NfL) as
fixed effects as well as random subject effects to account for correlations from repeated
measures in the same subjects; Model-2 is the same as Model 1, but without variables
involving NfL. 1000 simulations were performed. For each simulation, treatment group
assignment was simulated at random, which assigned half of the patients with ≥3
ALSFRS-R scores (n=53) to the treatment group and the remaining to the control group.
With �� and �� representing the estimated standard errors of time x group effects in
Model-1 and Model-2, respectively, the percent reduction in sample size is proportional
to the reduction in variance given by 100 � �1 � ����/��
�.
We also explored the utility of using serum NfL as a biomarker to detect
pharmacodynamic effect in a phase-2 clinical trial. Sample size estimations for such a
trial required us to define a difference (or change) in serum neurofilament that might be
regarded as clinically meaningful and which a future trial would need to be powered to
detect. First, we considered the difference in serum NfL between fast and slow
progressing patients (ALSFRS-R rate of decline >1 and <0.5 point/month, respectively),
reasoning that this difference in neurofilament could be considered meaningful since it
corresponds to a clinically important difference in rates of disease progression. We also
considered an alternative approach, in which we defined a meaningful change based on
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
9
9
a slope that is more negative (i.e. steeper decline) than the lower bound of the 95%
confidence interval of NfL trajectories in untreated patients. All analyses were
performed using SAS (version 9.4) and R software (https://www.r-project.org/).
Results
Study Population
A total of 259 patients (228 ALS, 11 PMA, and 20 PLS), with a total of 614 person-visits,
were included in this study (Table-1). For evaluation of the pharmacodynamic potential
of neurofilament measurements, analysis was restricted to the subset of 106 ALS, 3
PMA, and 4 PLS patients with at least 3 longitudinal biofluid collections (Table e1).
Analytic Characteristics of Neurofilament Assays in Serum and CSF
In CSF, both replicates were very infrequently below assay LLDs (i.e. undetectable);
this was true for both NfL and pNfH and in all assays. For serum, undetectable values
were infrequent using the Simoa NfL and pNfH assays, but were encountered in ~4%
and ~10% respectively of samples assayed using the Iron Horse and Euroimmun pNfH
platforms (Table-2). Agreement between technical replicates was high for all assays
except for CSF analysis using the Euroimmun assay (Table-2b). The mean CVs for NfL
were ~3% in serum and CSF; and for pNfH were ~4-5% and ~2-3% for all assays in
serum and CSF, respectively. We selected the Simoa platform for subsequent analyses,
given that it had the fewest undetectable values, and good replicate agreement (low
CV) in both serum and CSF.
Baseline Neurofilament Concentrations
Initial serum concentrations of NfL were higher in the ALS group compared to those with
PMA and PLS (p<0.001; Table-1). In the ALS group, higher serum NfL was associated
with older age, higher ∆FRS, female, the presence of a C9ORF72 repeat expansion,
and bulbar symptom onset (all p<0.001), but not with baseline ALSFRS-R. Higher CSF
NfL levels were similarly associated with older age, higher ∆FRS, and female gender,
although not reaching statistical significance (possibly due to smaller sample size).
Higher baseline serum pNfH among ALS patients was associated with older age
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
10
10
(p=0.046), higher ∆FRS (p<0.001), female gender (p=0.002), and bulbar onset disease
(p=0.035), but not C9ORF72 repeat expansion status or baseline ALSFRS-R. Higher
CSF pNfH levels were also associated with a higher ∆FRS, although not reaching
statistical significance.
Correlation between Serum and CSF Neurofilament Concentrations
While CSF could more directly reflect CNS pathophysiology, serum is more easily
accessible. Therefore, we examined for each assay how well the value of serum
measurements could serve as a proxy for CSF measurements. For pNfH, the serum-
CSF correlations were comparable among the 3 assays but weak (r=0.16 to 0.22); the
serum-CSF correlation was, however, much stronger for NfL (r=0.62, Simoa platform
only).
Potential Prognostic Utility of Baseline Neurofilament Concentrations
The average (±SD) rate of decline of the ALSFRS-R was 0.65 (±0.65) points per month.
Multivariate regression analysis, with the ALSFRS-R slope as the outcome, considering
potential clinical predictors of prognosis (age, gender, C9ORF72 status, site of disease
onset and ∆FRS) as well as baseline serum NfL and pNfH, identified only ∆FRS as a
meaningful clinical predictor. For every 1-point increase in the ∆FRS (pre-slope), the
ALSFRS-R rate of decline is worsened by an additional 0.46 points/month (p=0.004)
(Table-3). The addition of baseline serum NfL to the model shows that this biomarker
adds prognostic value that is not already explained by known clinical predictors. For
every 1-point increase in log serum NfL level, the ALSFRS-R rate of decline is
worsened by an additional 0.35 points/month (p<0.001). By contrast, baseline serum
pNfH provided little additional prognostic value (p=0.11). When both NfL and pNfH were
jointly considered, the results showed that, together, these biomarkers added
significantly more prognostic values independent of clinical predictors (p=0.002,
likelihood ratio test); this result, however, was primarily driven by the effect of NfL.
In univariate survival analyses, the presence of C9ORF72 repeat expansion, higher
∆FRS at baseline, and higher baseline serum NfL levels, were each associated with
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
11
11
worse prognosis (shorter survival), but baseline serum pNfH levels were not (Figure-1).
In a multivariate Cox proportional hazards model that considered age, gender,
C9ORF72 status, site of disease onset, baseline ALSFRS-R, ∆FRS, and baseline
serum levels of NfL and pNfH, only age, baseline ALSFRS-R, ∆FRS, and serum NfL
were significant predictors of survival (Table-3).
Potential Pharmacodynamic Utility of Longitudinal Neurofilament Concentrations
To ready the use of serum NfL and pNfH as potential pharmacodynamic biomarkers of
treatment effect in future trials, we estimated their longitudinal trajectories (and
variability of slopes). For serum NfL, the average slope was 0.011 log units/month (95%
confidence intervals [CI]: -0.054, 0.076). For serum pNfH, the average slope was 0.006
log units/month (95% CI: -0.063, 0.084) (Figure-2). We also quantified the magnitude
change in biomarker concentration on the raw scale compared to baseline (Figure-e1).
For serum NfL, the 90th and 95th percentiles of the maximum absolute changes in
concentration from baseline are 15 and 22pg/ml respectively. For serum pNfH, the 90th
and 95th percentiles of the maximum absolute changes in concentration from baseline
are 146 and 196pg/ml respectively.
Utility of Neurofilament Biomarkers in Reducing Sample Size for Future Clinical Trials
Since baseline NfL level is predictive of ALSFRS-R trajectories, we hypothesized that
the number of patients needed for a trial would be reduced when baseline NfL levels, as
a prognostic biomarker, are considered. This may be done by including serum NfL as a
covariate in the linear mixed models typically used to examine ALSFRS-R slopes. Using
a simulation study similar to that described by Kuffner23, we found that the sample size
for an ALS trial would be reduced by about 8.2% with the addition of baseline NfL
measurements.
Furthermore, we estimated the sample size required for a future phase-2 trial with
serum NfL as a pharmacodynamic biomarker. We considered 2 different outcome
measures and what a clinically meaningful treatment difference may be for each: First,
assuming that an experimental therapeutic would need to change the slope of serum
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
12
12
NfL from our observed average in the untreated state (0.011 log units/month) to less
than the lower bound of the 95% confidence interval (-0.054 log units/month) and
thereby an estimated treatment difference of 0.065 log units/month, with an estimated
standard deviation of 0.048 log units/month, a total sample size of N=26 (13 treatment,
13 placebo) would provide 90% power to detect such a treatment difference, using a
two-sample t-test with 5% significance level. Alternatively, one may wish to achieve a
clinically meaningful difference in NfL levels (rather than trajectories). Based on an
estimated treatment difference of 0.67 log(pg/ml)—which was our observed difference
in baseline log(NfL) levels between fast progressors (ALSFRS-R decline of >1
point/month) and slow progressors (ALSFRS-R decline of <0.5 points/month), which
were 3.10 and 2.43 log(pg/ml), respectively—and a standard deviation of ~0.81
log(pg/ml), a total sample size of N=64 (32 treatment, 32 placebo) would provide 90%
power to detect such a treatment difference, using a two sample t-test with 5%
significance level.
For comparison, we also similarly estimated the required sample size for a clinically
meaningful 20-30% reduction in ALSFRS-R slope, a common outcome measure
employed in ALS clinical trials 24. Based on our observed average ALSFRS-R slope of -
0.648 points/month and an estimated standard deviation of 0.65 points/month, a total of
N=1054 or N=470 patients would be needed respectively to provide power for detecting
a 20% or 30% reduction in ALSFRS-R slopes.
Discussion
We have undertaken a head-to-head comparison of three different pNfH assays
(performed by two different CROs) in serum and CSF, as well as an evaluation of the
prognostic and potential pharmacodynamic utility of serum NfL and pNfH. For pNfH, the
Simoa assay had the best sensitivity (fewest values below the assay LLD) and
reproducibility (lowest CV between technical replicates in both serum and CSF). Using
the Simoa assay, we also found a much stronger correlation between serum and CSF
for NfL than for pNfH. The correlations between serum and CSF in this study are
substantially lower than prior published reports 11, 20, 25. One possible explanation for this
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
13
13
discrepancy is that we restricted our analyses to baseline data. As discussed by Bland
and Altman, when multiple measurements taken from the same subjects are used to
compute correlation coefficients, the result can appear artificially high because
variability between measurements on the same subject are not properly accounted for
and the independence assumption of correlation is violated 22.
We also found that serum NfL, but not serum pNfH, is prognostic of future ALSFRS-R
decline and survival duration, providing information that is not captured by readily
available clinical predictors. While absolute values can vary between patients, serum
NfL and pNfH levels remain largely stable in each subject over time. This may portend
clinical utility as pharmacodynamic biomarkers if there are detectable changes in levels
following exposure to an experimental therapeutic. This expectation is supported by
recent studies of Nusinersen for spinal muscular atrophy (SMA) 26, 27, though definitive
proof in adult motor neuron diseases would necessarily require an effective therapeutic
to test. Moreover, we have illustrated the ways in which use of serum NfL in a phase-2
trial might permit a reduction in sample size. Incorporation of serum NfL as a covariate
in a trial that uses the ALSFRS-R rate of decline as the primary outcome measure
would yield a modest benefit (~8% reduction in sample size), but use of NfL as a
pharmacodynamic biomarker might offer meaningful sample size savings compared to
more traditional phase-2 studies in which changes in the ALSFRS-R are used as the
primary outcome measure.
The main limitation of this study is that the population may not be adequately
representative of ALS patients who would ordinarily be enrolled in a clinical trial.
Patients included in the current study were skewed towards those with more slowly
progressive disease (evidenced by an ALSFRS-R slope of -0.65 points/month).
Achieving better representation of a trial-like population is the goal of an ongoing study.
A second limitation is that we have insufficient data to reliably comment on the value of
CSF NfL and pNfH as either prognostic or pharmacodynamic biomarkers.
Notwithstanding these considerations, this was a large multi-center study with
prospective, systematic follow-up and strict SOPs for sample collection, processing and
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
14
14
storage. All assays were performed by independent CROs blinded to clinical data, with
comparative data generated across assay platforms, CROs and neurofilament types.
These results support two conclusions. First, serum NfL, but not serum pNfH, may be
considered a clinically validated prognostic biomarker. Second, both serum NfL and
pNfH may have value as potential pharmacodynamic biomarkers, and both should be
incorporated into ongoing and future phase-2 trials; the actual pharmacodynamic utility
of these biomarkers, however, will only become apparent once we have truly effective
treatments for ALS.
Data Sharing Statement
De-identified data will be shared immediately following publication.
Appendix 1 Authors Name Location Role Contribution
Michael Benatar, MD, PhD University of Miami Author
Study conceptualization and design; data acquisition; study oversight; statistical analysis and results interpretation; drafting of the manuscript for intellectual content; critical review and revision of manuscript
Lanyu Zhang, MS University of Miami Author Statistical analysis; review of
manuscript for intellectual content Lily Wang, PhD
University of Miami Author Statistical analysis; review of manuscript for intellectual content
Volkan Granit, MD University of Miami Author Data acquisition; critical review of
manuscript for intellectual content
Jeffrey Statland, MD
United States, Kansas University Medical Center
Author Data acquisition; critical review of manuscript for intellectual content
Richard Barohn, MD
United States, Kansas University Medical Center
Author Data acquisition; critical review of manuscript for intellectual content
Andrea Swenson, MD University of Iowa Author Data acquisition; critical review of
manuscript for intellectual content John Ravits, MD
University of California San Diego Author
Data acquisition; critical review of manuscript for intellectual content
Carlayne Jackson, University of Texas Author Data acquisition; critical review of
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
15
15
MD Health Sciences Center, San Antonio
manuscript for intellectual content
Ted M Burns, MD University of Virginia Author Data acquisition; critical review of
manuscript for intellectual content
Jaya Trivedi, MD
University of Texas Southwestern Medical Center
Author Data acquisition; critical review of manuscript for intellectual content
Erik P Pioro, MD, PhD Cleveland Clinic Author Data acquisition; critical review of
manuscript for intellectual content James Caress, MD
Wake Forest School of Medicine
Author Data acquisition; critical review of manuscript for intellectual content
Jonathan Katz, MD
California Pacific Medical Center Author Data acquisition; critical review of
manuscript for intellectual content
Jacob L McCauley, PhD
John P Hussman Institute for Human Genomics
Author Data acquisition; critical review of manuscript for intellectual content
Rosa Rademakers, PhD
Mayo Clinic, Jacksonville Author Data acquisition; critical review of
manuscript for intellectual content
Andrea Malaspina, MD, PhD
Blizard Institute, Queen Mary University of London
Author Results interpretation; critical review of manuscript for intellectual content
Lyle W Ostrow, MD, PhD
Johns Hopkins University Author
Study design; results interpretation; critical review of manuscript for intellectual content
Joanne Wuu, ScM University of Miami, Author
Study conceptualization and design; data acquisition; study oversight; statistical analysis and results interpretation; drafting of the manuscript for intellectual content; critical review and revision of manuscript
Appendix 2
Co-Investigators (CReATe Consortium)
Name Location Role Contribution
Sumaira Hussain University of Miami, USA Project Manager
Regulatory lead; multi-center oversight and project management
Anne Cooley University of Miami, USA Associate Project
& Data Manager
Multi-center oversight and project management; data query resolution
Yindi Li University of Miami, USA
Statistical programmer Data management
Marielle Wallace University of Miami, USA Data Manager Data management
Julie Steele University of Miami, Outcomes Training and certification of
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
16
16
USA Coordinator site coordinators in outcome measures
Jessica P. Hernandez
University of Miami, USA
Research Coordinator
Visit coordination and data collection
Jessica Medina University of Miami, USA
Research Coordinator
Visit coordination and data collection
Maria Elena Paredes
University of Miami, USA
Research Coordinator
Visit coordination and data collection
Ashley Manso University of Miami, USA
Research Coordinator Data collection and review
Natalia Ravelo University of Miami, USA
Research Coordinator
Visit coordination and data collection
Wendy Levy University of Miami, USA
Research Coordinator
Visit coordination and data collection
Patrice Whitehead John P Hussman Institute for Human Genomics
Biorepository Technical Director
CReATe Biorepository laboratory oversight
Stephan Zuchner, MD, PhD
University of Miami, USA Co-investigator
CReATe Consortium Genetic Working Group member
Mamatha Pasnoor, MD
University of Kansas Medical Center, USA Sub-investigator Data collection
Omar Jawdat, MD University of Kansas Medical Center, USA Sub-investigator Data collection
Duaa Jabari, MD University of Kansas Medical Center, USA Sub-investigator Data collection
Constantine Farmakidis, MD
University of Kansas Medical Center, USA Sub-investigator Data collection
Melanie Glenn, MD
University of Kansas Medical Center, USA Sub-investigator Data collection
Mazen M Dimachkie, MD
University of Kansas Medical Center, USA Sub-investigator Data collection
Laura Herbelin University of Kansas Medical Center, USA
Site Lead Coordinator
Visit coordination and data collection
Hellen Tanui University of Kansas Medical Center, USA
Research Coordinator
Visit coordination and data collection
Sherri Anderson University of Kansas Medical Center, USA
Research Coordinator
Visit coordination and data collection
Michaela Walker University of Kansas Medical Center, USA
Research Coordinator
Visit coordination and data collection
Tina Liu University of Kansas Medical Center, USA
Research Coordinator
Visit coordination and data collection
Ayla McCally University of Kansas Medical Center, USA
Research Coordinator
Visit coordination and data collection
Andrew Heim University of Kansas Medical Center, USA
Research Coordinator
Visit coordination and data collection
Melissa Currence University of Kansas Medical Center, USA
Research Coordinator
Visit coordination and data collection
Yolanda Harness University of Kansas Medical Center, USA
Research Coordinator
Visit coordination and data collection
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
17
17
Jeri Sieren University of Iowa, USA
Research Coordinator
Visit coordination and data collection
Emilee Gibson University of Iowa, USA
Research Coordinator
Visit coordination and data collection
Gil Gutierrez University of California San Diego, USA
Research Coordinator Visit coordination and data
collection
Danielle Bussey University of California San Diego, USA
Research Coordinator Visit coordination and data
collection
Rose Previte University of California San Diego, USA
Research Coordinator Visit coordination and data
collection
Pamella Kittrell
University of Texas Health Science Center at San Antonio, USA
Research Coordinator Visit coordination and data
collection
Amruta Joshi University of Virginia, USA
Research Coordinator
Visit coordination and data collection
Amy Conger UT Southwestern Medical Center, USA
Research Coordinator
Visit coordination and data collection
Debbie Hastings Cleveland Clinic, USA
Research Coordinator
Visit coordination and data collection
Irys Caristo Cleveland Clinic, USA
Research Coordinator
Visit coordination and data collection
Mozhdeh Marandi Wake Forest School of Medicine, USA
Research Coordinator
Visit coordination and data collection
Simon Carty California Pacific Medical Center, USA
Research Coordinator
Visit coordination and data collection
J Paul Taylor, MD, PhD
St. Jude Children’s Research Hospital, USA
Co-investigator CReATe Consortium genomics lead
Gang Wu, PhD
St. Jude Children’s Research Hospital, USA Co-investigator
CReATe Consortium genomic data analysis and Genetic Working Group member
Evadnie Rampersaud, PhD
St. Jude Children’s Research Hospital, USA Co-investigator
CReATe Consortium genomic data analysis and Genetic Working Group member
Rebecca Schule, MD, PhD
University of Tubingen, Germany Co-investigator
CReATe Consortium Genetic Working Group member
Marka van Blitterswijk, MD, PhD
Mayo Clinic Jacksonville, USA Co-investigator
CReATe Consortium Genetic Working Group member
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
18
18
Acknowledgements
We would like to acknowledge Joaquin del Cueto (Project Manager for Research
Support, University of Miami); Dr. Lucie Bruijn (ALS Association Chief Scientist); Drs.
Amelie Gubitz and Robin Conwit (NINDS Program Directors); Dr. Tiina Urv (NCATS
Program Officer); Dr. Manish Raisinghani (CEO, Target ALS Foundation); and the Rare
Diseases Clinical Research Network (RDCRN) Data Management and Coordinating
Center (DMCC) at University of South Florida (PI: Dr. Jeffrey Krischer; regulatory
specialists: Julie Martin and Kaleena Dezsi; and project managers: Callyn Kirk, Devon
Rizzo, and Kristina Bowles). We would also like to thank Drs. Zaven Kaprielian
(Amgen), Danielle Graham and Toby Ferguson (Biogen), Joe Lewcock (Denali), Felix
Yeh (Genentech), Sophie Batteur Parmentier (Merck), Bryan Hill (Mitsubishi Tanabe
Pharma), Dave Frendewey (Regeneron), Aarti Sharma (Regeneron), James Dodge
(Sanofi), Laure Rosen (Takeda), Ian Reynolds and Neta Zach (Teva), Kelly Dakin
(FBRI), Nadine Tatton (AFTD), Amanda Haidet-Phillips (MDA), Robert Bowser (Barrow
Neurological Institute) and Leonard Petrucelli (Mayo Clinic Jacksonville) for helpful
discussions around selecting biomarker candidates and contract research
organizations. We would also like to acknowledge the staff and support of the John P.
Hussman Institute for Human Genomics Biorepository Facility at the University of Miami
School of Medicine. Most importantly, we are indebted to patients who participated in
the CReATe Phenotype-Genotype-Biomarker (PGB) study for their altruism as well as
their commitment and contribution to advancing therapeutic development for ALS and
related disorders.
Funding
The CReATe Consortium (U54 NS090291) is part of the Rare Diseases Clinical
Research Network (RDCRN), an initiative of the Office of Rare Diseases Research
(ORDR), National Center for Advancing Translational Sciences (NCATS). CReATe is
funded through collaboration between NCATS and the National Institute of Neurological
Disorders and Stroke (NINDS). CReATe is also supported by a Clinical Trial Readiness
Grant from NIH (U01NS107027). Supplementary support for the CReATe Biorepository
was provided by the ALS Association (Grant ID 16-TACL-242). Target ALS provided
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
19
19
support to Quanterix and Iron Horse to cover assay costs as well as through a grant to
Joanne Wuu. This work was also supported by CTSA grants (UL1TR002366 and
UL1TR000001) from NCATS and awarded to the University of Kansas.
References
1. Benatar M. ALS Therapy Development: Challenges and Opportunities. Journal of
Neuromuscular Diseases 2016;3:S10.
2. van den Berg LH, Sorenson E, Gronseth G, et al. Revised Airlie House consensus
guidelines for design and implementation of ALS clinical trials. Neurology
2019;92:e1610-e1623.
3. Benatar M, Boylan K, Jeromin A, et al. ALS biomarkers for therapy development:
State of the field and future directions. Muscle & nerve 2016;53:169-182.
4. Verde F, Steinacker P, Weishaupt JH, et al. Neurofilament light chain in serum for
the diagnosis of amyotrophic lateral sclerosis. J Neurol Neurosurg Psychiatry
2019;90:157-164.
5. Rossi D, Volanti P, Brambilla L, Colletti T, Spataro R, La Bella V. CSF
neurofilament proteins as diagnostic and prognostic biomarkers for amyotrophic
lateral sclerosis. Journal of neurology 2018;265:510-521.
6. Gaiani A, Martinelli I, Bello L, et al. Diagnostic and Prognostic Biomarkers in
Amyotrophic Lateral Sclerosis: Neurofilament Light Chain Levels in Definite
Subtypes of Disease. JAMA neurology 2017;74:525-532.
7. Steinacker P, Huss A, Mayer B, et al. Diagnostic and prognostic significance of
neurofilament light chain NF-L, but not progranulin and S100B, in the course of
amyotrophic lateral sclerosis: Data from the German MND-net. Amyotroph Lateral
Scler Frontotemporal Degener 2017;18:112-119.
8. Oeckl P, Jardel C, Salachas F, et al. Multicenter validation of CSF neurofilaments
as diagnostic biomarkers for ALS. Amyotroph Lateral Scler Frontotemporal
Degener 2016;17:404-413.
9. Boylan KB, Glass JD, Crook JE, et al. Phosphorylated neurofilament heavy subunit
(pNF-H) in peripheral blood and CSF as a potential prognostic biomarker in
amyotrophic lateral sclerosis. J Neurol Neurosurg Psychiatry 2012.
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
20
20
10. Gendron TF, Group CONS, Daughrity LM, et al. Phosphorylated neurofilament
heavy chain: A biomarker of survival for C9ORF72-associated amyotrophic lateral
sclerosis. Annals of neurology 2017;82:139-146.
11. Lu CH, Macdonald-Wallis C, Gray E, et al. Neurofilament light chain: A prognostic
biomarker in amyotrophic lateral sclerosis. Neurology 2015;84:2247-2257.
12. Lu CH, Petzold A, Topping J, et al. Plasma neurofilament heavy chain levels and
disease progression in amyotrophic lateral sclerosis: insights from a longitudinal
study. J Neurol Neurosurg Psychiatry 2015;86:565-573.
13. McCombe PA, Pfluger C, Singh P, Lim CY, Airey C, Henderson RD. Serial
measurements of phosphorylated neurofilament-heavy in the serum of subjects with
amyotrophic lateral sclerosis. Journal of the neurological sciences 2015;353:122-
129.
14. Poesen K, De Schaepdryver M, Stubendorff B, et al. Neurofilament markers for
ALS correlate with extent of upper and lower motor neuron disease. Neurology
2017;88:2302-2309.
15. Boylan K, Yang C, Crook J, et al. Immunoreactivity of the phosphorylated axonal
neurofilament H subunit (pNF-H) in blood of ALS model rodents and ALS patients:
Evaluation of blood pNF-H as a potential ALS biomarker. J Neurochem 2009.
16. Lu CH, Petzold A, Kalmar B, Dick J, Malaspina A, Greensmith L. Plasma
neurofilament heavy chain levels correlate to markers of late stage disease
progression and treatment response in SOD1(G93A) mice that model ALS. PLoS
One 2012;7:e40998.
17. Kuhle J, Barro C, Andreasson U, et al. Comparison of three analytical platforms for
quantification of the neurofilament light chain in blood samples: ELISA,
electrochemiluminescence immunoassay and Simoa. Clinical chemistry and
laboratory medicine 2016;54:1655-1661.
18. Quanterix. Simoa NF-light [online]. Available at:
https://www.quanterix.com/sites/default/files/assays/Simoa_NF-
light_Data_Sheet_HD-1.pdf.
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
21
21
19. Quanterix. Simoa pNF_heavy [online]. Available at:
https://www.quanterix.com/sites/default/files/assays/Simoa%20pNF-
heavy%20Discovery%20data%20sheet.pdf.
20. De Schaepdryver M, Jeromin A, Gille B, et al. Comparison of elevated
phosphorylated neurofilament heavy chains in serum and cerebrospinal fluid of
patients with amyotrophic lateral sclerosis. J Neurol Neurosurg Psychiatry
2018;89:367-373.
21. Kimura F, Fujimura C, Ishida S, et al. Progression rate of ALSFRS-R at time of
diagnosis predicts survival time in ALS. Neurology 2006;66:265-267.
22. Bland JM, Altman DG. Statistical methods for assessing agreement between two
methods of clinical measurement. Lancet 1986;1:307-310.
23. Kuffner R, Zach N, Norel R, et al. Crowdsourced analysis of clinical trial data to
predict amyotrophic lateral sclerosis progression. Nat Biotechnol 2015;33:51-57.
24. Castrillo-Viguera C, Grasso DL, Simpson E, Shefner J, Cudkowicz ME. Clinical
significance in the change of decline in ALSFRS-R. Amyotroph Lateral Scler
2010;11:178-180.
25. Weydt P, Oeckl P, Huss A, et al. Neurofilament levels as biomarkers in
asymptomatic and symptomatic familial amyotrophic lateral sclerosis. Annals of
neurology 2016;79:152-158.
26. Winter B, Guenther R, Ludolph AC, Hermann A, Otto M, Wurster CD.
Neurofilaments and tau in CSF in an infant with SMA type 1 treated with
nusinersen. J Neurol Neurosurg Psychiatry 2019.
27. Darras BT, Crawford TO, Finkel RS, et al. Neurofilament as a potential biomarker
for spinal muscular atrophy. Annals of clinical and translational neurology
2019;6:932-944.
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
22
22
Figure Legends
Figure 1. Kaplan-Meier survival curves, showing the prognostic value of (A)
C9ORF72 repeat expansion, (B) ∆FRS (dichotomized at the median, 0.62points/month),
(C) baseline serum NfL (dichotomized at the median, 17pg/ml), and (D) baseline serum
pNfH (dichotomized at the median, 67pg/ml). The presence of a C9ORF72 repeat
expansion, higher ∆FRS, higher baseline serum NfL, and higher baseline serum pNfH
are shown in red.
Figure 2. Longitudinal trajectories of serum neurofilaments: (A) spaghetti plot of
log-transformed neurofilament light level (NfL), (B) spaghetti plot of log-transformed
phosphorylated neurofilament heavy level (pNfH), and (C) boxplot of the distribution of
NfL and pNfH slopes. Slope estimates were obtained from a linear model with log-
transformed neurofilament level as the outcome, and time as the independent variable.
Supplemental Figure 1. Change in serum neurofilaments over time, as compared
to baseline level: (A) Change from baseline serum neurofilament light (pg/ml), and (B)
change from baseline in serum phosphorylated neurofilament heavy (pg/ml).
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
23
23
Tables Table 1. Study Participant Characteristics ALS
(N=228) PMA
(N=11) PLS
(N=20)
Baseline age (years) Mean ± SD 60.1 ± 11.8 56.8 ± 11.6 58.8 ± 11.1
Male N (%) 133 (58%) 8 (73%) 9 (45%)
Genotype: C9ORF72 Other Unknown
N (%)
27 (12%) 10 ( 4%) 191 (84%)
0 0
11 (100%)
0 0
20 (100%) Site of onset:
Bulbar only Limb only Other region(s) only Multiple regions Unknown
N (%)
41 (18%) 151 (66%) 4 ( 2%) 24 (11%) 8 ( 4%)
0
11 (100%) 0 0 0
6 (30%) 9 (45%)
0 4 (20%) 1 ( 5%)
Years from onset of weakness to baseline Median (IQR) 1.8 (1.0 – 3.2) 2.4 (1.7 – 9.0) 7.9 (3.8 – 11.6)
Years from diagnosis to baseline Median (IQR) 0.8 (0.3 – 1.7) 1.0 (0.6 – 4.3) 2.5 (1.5 – 5.5)
Baseline ALSFRS-R Mean (± SD) 34.7 ± 7.5 33.5 ± 8.5 35.2 ± 8.1
Baseline ∆FRS Mean (± SD) 0.62 ± 0.48 0.34 ± 0.36 0.17 ± 0.12
Baseline serum NfL (pg/ml) a Median (range) 17 (2, 369) b 7 (3, 26) 10 (2, 54)
Baseline serum pNfH (pg/ml) a Median (range) 67 (1, 976) b 27 (2, 101) 30 (1, 328)
Baseline CSF NfL (pg/ml) a Median (range) 107 (12, 269) c N/A N/A
Baseline CSF pNfH (pg/ml) a Median (range) 103 (16, 216) c N/A N/A
ALSFRS-R slope d Mean (± SD) -0.65 ± 0.65 -0.57 ± 0.14 -0.09 ± 0.46
Survival time (months) Median (IQR) 25.9 (14.4 - 43.3)
24.6 (24.6 - 33.7) Not estimable e
Baseline = first visit at which serum sample was available (with or without contemporaneous CSF collection). IQR = inter-quartile range. N/A = not available. a Neurofilament data generated through Simoa assays b Baseline serum data not available for 3 ALS participants c Baseline CSF data only available for 29 ALS participants d Calculated from the 106 ALS, 3 PMA and 4 PLS participants in the longitudinal subset (i.e. participants with at least
three neurofilament measurements) e Median survival time could not be reliably estimated because only 2 PLS patients reached survival endpoint during study follow-up
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
24
24
Table 2a. Analytical Characteristics of Neurofilament Light: Simoa Assay Simoa
Serum CSF
Total number of samples 614 78
Number of samples with value(s) below LLD:
In both replicates 0 0 In only one replicate 4 1
Difference (in pg/ml) between replicates, mean (95% CI) a
-0.05 (-4.0, 3.9)
-0.03 (-11.8, 11.7)
Coefficient of variation (CV) b, mean 3.2 2.9
Number (proportion) of samples with CV > 10
12 (2.0%)
1 (1.3%)
LLD = lower limit of detection a Difference = replicate 2 – replicate 1. The 95% CI of this difference is also known as the limits of agreement. b CV = 100 x (standard deviation of replicates 1 and 2) / (mean of replicates 1 and 2). Table 2b. Analytical Characteristics of Phosphorylated Neurofilament Heavy (pNfH): Simoa Assay, Iron Horse Assay, and Euroimmun Assay
Simoa Iron Horse Euroimmun
Serum CSF Serum CSF Serum CSF
Total number of samples 614 78 614 78 614 78
Number of samples with value(s) below LLD:
In both replicates 0 1 26 0 64 0 In only one replicate 15 2 36 0 73 0
Difference (in pg/ml) between replicates, mean (95% CI) a
-0.4 (-26, 25)
1.0 (-7, 9)
-0.8 (-16, 14)
-4.9 (-90, 80)
-0.02 (-38, 38)
67.9 (-163, 298)
Coefficient of variation (CV) b, mean 4.7 2.1 4.0 2.7 5.3 3.0
Number (proportion) of samples with CV > 10
51 (8.3%)
0 (0.0%)
63 (10.3%
)
0 (0.0%)
88 (14.3%)
0 (0.0%)
LLD = lower limit of detection a Difference = replicate 2 – replicate 1. The 95% CI of this difference is also known as the limits of agreement. b CV = 100 x (standard deviation of replicates 1 and 2) / (mean of replicates 1 and 2).
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Benatar et al
25
25
Table 3. Impact of Clinical and Neurofilament Characteristics on ALSFRS-R Rate of Decline and Survival
Covariates
Impact on ALSFRS-R Slope a
Impact on Survival b
Estimate (95% CI) p-value Hazard Ratio
(95% CI) p-value
Baseline age (years) 0.002
(-0.007, 0.012)
0.67 1.02 (1.00, 1.05)
0.049
Male -0.09 (-0.33, 0.16) 0.49 1.37
(0.84, 2.23) 0.21
C9ORF72 HREM -0.18 (-0.60, 0.24)
0.40 1.70 (0.91, 3.18)
0.10
Site of Onset: Bulbar -0.01 (-0.43, 0.40) 0.94 1.10
(0.53, 2.29) 0.80
Site of Onset: Limb 0.03 (-0.44, 0.49) 0.92 1.04
(0.48, 2.27) 0.91
Baseline ALSFRS-R N/A N/A 0.96 (0.93, 0.99)
0.02
Baseline ∆FRS (points/month) -0.46 (-0.76, -0.15)
0.004 1.67 (1.09, 2.57)
0.02
Baseline log(NfL) (log(pg/ml)) -0.35 (-0.55, -0.15)
< 0.001 2.12 (1.39, 3.23)
< 0.001
Baseline log(pNfH) (log(pg/ml)) 0.08 (-0.02, 0.18)
0.11 1.01 (0.84, 1.22)
0.88
CI = confidence interval. N/A = not applicable. a Based on multivariate regression analysis, with the ALSFRS-R slope (i.e. ALSFRS-R rate of decline) as outcome b Based on Cox proportional hazards model, with tracheostomy- and permanent assisted ventilation (PAV)-free survival as outcome
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
. CC-BY-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not certified by peer review)
(which wasThe copyright holder for this preprint this version posted August 3, 2019. .https://doi.org/10.1101/19002998doi: medRxiv preprint
Top Related